Wall mesh size high Re nummer flow through duct
 

Hi all,

I am modeling a pressure driven water flow through a square duct 3.2x3.2m.

The wall has an equivalent roughness of 3mm. Flow velocity is 5m/s. I am using the k-omega turbulence model.

As I understand it I cannot fully resolve the flow near the wall, since this would require a very fine mesh with this Reynolds number.

Now k-omega model uses the automated wall function to model the flow near the wall.

My question is: what mesh size is in this case appropriate near the wall?

An order of magnitude answer from experience is already useful.

The most complete answer is: Do a sensitivity study and find out. Different simulations have different requirements, so to find out for your case test a range of near wall mesh sizes and find out what you require.

So unless I know a lot more about your physics I cannot give a recommended y+.

I note you are using the k-omega turbulence model. Why have you chosen that one? The usual recommended turbulence model is SST.

Hi Glenn,

Thanks for the reply. To give some more information I added two image.

1. A drawing of the civil structure I am working on. It consists of 6 channels each with a cross section 3.2x3.2m. Most occurring water level difference over the structure is 1.5m (7m versus 5.5m). This results in velocities of ~4.5m/s. The drawing also indicates where we did measurements with a current profiler.
2. A result of my CFX model and the measurements over the height of the channel. The measurement are the dots. I trust they are correct.

The modelled profile is more curved then the measured one. Could this be due to:
- too high wall roughness (3mm in this case)
- mesh (using a 0.1m hex mesh with inflation to the wall up to 2mm)
- turbulence model (using k-omega. I understood that this works well in confined flows)
- something else

I am on a learning curve, still trying to fully understand yplus values and turbulence modelling.

Any thoughts are appreciated

I would recommend you use the SST turbulence model unless you have a good reason not to. It includes all the physics of the k-omega model and has a lot more as well.

Your results are pretty close, but it seems you want to get even better so you can resolve the details of the flow (ie the curvature int he velocity profile). When you start looking for highly accurate CFD results you need to be careful to verify and validate your model. Here is a starting point for that: https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

In your case I would suspect mesh resolution will be the key issue to control simulation accuracy.